Is a digital elevation map the same as an aerial photograph?

No. It stores elevation-related values, while a photograph records appearance. They can be displayed together but represent different information.

Does DEM always mean bare earth?

Usage varies. Some providers use that meaning, while broader terminology exists. Check the dataset specification.

Can a grid represent a bridge and the ground beneath it at the same horizontal location?

A single-valued elevation grid ordinarily stores one surface height there, so additional representation is needed for both surfaces.

AI Glossary / Computer vision and multimodal AI / Beginner

Digital elevation map

Definition

A digital elevation map is a digital representation of surface elevation across geographic locations, commonly stored as a grid of height values. The closely related term digital elevation model, or DEM, is more standard in geospatial work. Its interpretation depends on the represented surface, spatial resolution, vertical reference, and measurement quality.

What it means

Each grid location records an elevation relative to a stated reference, rather than ordinary image brightness. Software can use those values to show terrain, calculate slopes, or support other geographic analyses.

Terminology varies. USGS commonly describes a DEM as bare-earth terrain excluding buildings and vegetation, while some literature uses DEM more broadly and distinguishes digital terrain models from digital surface models. The dataset's specification is therefore more reliable than its short label alone.

An elevation grid is usually a two-and-a-half-dimensional representation: it assigns one height to each horizontal location. It does not fully describe overhangs, tunnels, vertical walls, or several surfaces stacked above the same location.

How it works

Collect elevation evidence from a suitable source, such as lidar, photogrammetry, radar-derived measurements, or existing surveying data. Process the measurements according to the surface the product is intended to represent.

For a bare-earth model, distinguish ground observations from buildings, vegetation, and other surface objects. Interpolate or aggregate onto the chosen grid where appropriate, documenting how gaps and uncertain areas are handled.

Attach a horizontal coordinate reference system, vertical datum, units, cell spacing, acquisition information, and quality metadata. These details determine whether two elevation datasets can be compared directly.

Use the grid for visualization or analysis, while checking whether its resolution and accuracy suit the question. A fine-looking shaded map can hide interpolation artifacts, stale measurements, or vertical offsets. Derived slopes and differences can amplify errors that were less obvious in the original elevation values.

Three practical examples

1. Landscape analysis: Researchers calculate broad slope patterns or compare landform geometry across an area.

2. Preliminary site review: A planning team examines terrain shape before commissioning the detailed measurements required for a specific design task.

3. Environmental monitoring: Comparable elevation datasets help study surface change, with explicit treatment of alignment and measurement uncertainty.

Think of it this way

Think of laying a regular mesh of height markers over a landscape. Each marker tells you the surface elevation at a defined horizontal location, and the collection forms a sampled description of the terrain.

The analogy has limits. Some marker values are interpolated rather than directly measured, and a single height per location cannot describe every part of a complex three-dimensional scene. The height reference also matters just as much as the numbers themselves.

Worked example

Suppose two neighboring grid-cell centers are ten meters apart horizontally, with elevations 100 and 102 meters in the same vertical datum. The rise is two meters over a ten-meter run.

The simple along-line slope is 2/10 = 0.2, or 20%. Its angle is arctan(0.2), approximately 11.3 degrees. This is a two-point directional calculation, not a complete terrain-slope algorithm using a surrounding neighborhood.

Now suppose a second dataset gives elevations 101 and 103 meters because its vertical reference differs by one meter. The local slope is unchanged, but direct subtraction from the first dataset would falsely suggest one meter of surface rise everywhere.

Before interpreting differences as landscape change, the team must reconcile references and evaluate uncertainty. Identical grid spacing alone does not make two elevation products comparable.

Illustrative case study

Fictional case: A conservation group compares two elevation grids to investigate changes along a riverbank. A first difference map shows an apparently uniform rise across stable ground as well as the bank.

The analysts check metadata and discover a vertical-reference mismatch. After appropriate reconciliation and alignment checks, they assess remaining changes against the datasets' uncertainty and inspect independent reference areas.

They also identify vegetation-related artifacts where one product represents canopy height and the other aims to represent bare earth. The project demonstrates why surface definition and datum must be resolved before a colorful difference map is treated as evidence of physical change.

Common misconceptions and mistakes

Do not confuse horizontal resolution with vertical accuracy. A one-meter grid does not automatically measure elevation to one-meter accuracy, or to any finer precision.

Do not assume all products exclude trees and buildings. Read the surface definition and processing specification.

Do not mix vertical datums or units when subtracting datasets. A reference mismatch can imitate terrain change.

Do not treat a hillshade image as the original elevation data. Shading is a visualization derived from heights and an assumed light direction, so its pixel brightness is not itself an elevation measurement.

Business use

Elevation maps provide a useful foundation for geographic visualization, terrain analysis, and early-stage planning. Their value depends on fitness for the intended scale and decision, not merely on visual detail.

Review acquisition date, coverage, surface type, reference systems, and documented uncertainty before use. For consequential engineering or environmental decisions, the elevation model is one input to an appropriately validated workflow. Preserve provenance and processing steps so later reviewers can distinguish measured change from resampling, filtering, or alignment effects.

Check your understanding

Two elevation grids have the same cell spacing but different vertical datums. Can their values be subtracted directly to measure terrain change?

Answer: No. Their vertical references must be reconciled and other comparability conditions checked before interpreting the difference.

Related terms

Digital elevation model
digital terrain model
digital surface model
lidar
photogrammetry
vertical datum
raster grid
hillshade

Learn next

Read the metadata for a small elevation dataset and identify its surface type, units, coordinate references, and uncertainty. Calculate one directional slope, then examine how a uniform vertical offset affects a change map without changing that slope.

Sources

USGS, What Is a Digital Elevation Model?: https://www.usgs.gov/faqs/what-a-digital-elevation-model-dem . The official explanation was read. The USGS publication summary Digital Elevation Models: Terminology and Definitions was consulted through indexed text for broader usage: https://www.usgs.gov/publications/digital-elevation-models-terminology-and-definitions . Calculations and the conservation case are original illustrations.

Previous: Digital divide Next: Digital geometry